2 * Misc image convertion routines
3 * Copyright (c) 2001, 2002 Fabrice Bellard.
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 #include "fastmemcpy.h"
25 /* XXX: totally non optimized */
27 static void yuv422_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
28 UINT8 *src, int width, int height)
33 for(y=0;y<height;y+=2) {
34 for(x=0;x<width;x+=2) {
44 for(x=0;x<width;x+=2) {
54 #define ONE_HALF (1 << (SCALEBITS - 1))
55 #define FIX(x) ((int) ((x) * (1L<<SCALEBITS) + 0.5))
57 static void rgb24_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
58 UINT8 *src, int width, int height)
60 int wrap, wrap3, x, y;
61 int r, g, b, r1, g1, b1;
67 for(y=0;y<height;y+=2) {
68 for(x=0;x<width;x+=2) {
75 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
76 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
83 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
84 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
94 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
95 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
102 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
103 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
105 cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 +
106 FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
107 cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 -
108 FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
120 static void rgba32_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
121 UINT8 *src, int width, int height)
123 int wrap, wrap4, x, y;
124 int r, g, b, r1, g1, b1;
130 for(y=0;y<height;y+=2) {
131 for(x=0;x<width;x+=2) {
138 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
139 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
146 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
147 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
157 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
158 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
165 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
166 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
168 cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 +
169 FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
170 cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 -
171 FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
183 static void bgr24_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
184 UINT8 *src, int width, int height)
186 int wrap, wrap3, x, y;
187 int r, g, b, r1, g1, b1;
193 for(y=0;y<height;y+=2) {
194 for(x=0;x<width;x+=2) {
201 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
202 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
209 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
210 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
220 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
221 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
228 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
229 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
231 cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 +
232 FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
233 cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 -
234 FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
246 static void bgra32_to_yuv420p(UINT8 *lum, UINT8 *cb, UINT8 *cr,
247 UINT8 *src, int width, int height)
249 int wrap, wrap4, x, y;
250 int r, g, b, r1, g1, b1;
256 for(y=0;y<height;y+=2) {
257 for(x=0;x<width;x+=2) {
264 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
265 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
272 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
273 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
283 lum[0] = (FIX(0.29900) * r + FIX(0.58700) * g +
284 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
291 lum[1] = (FIX(0.29900) * r + FIX(0.58700) * g +
292 FIX(0.11400) * b + ONE_HALF) >> SCALEBITS;
294 cb[0] = ((- FIX(0.16874) * r1 - FIX(0.33126) * g1 +
295 FIX(0.50000) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
296 cr[0] = ((FIX(0.50000) * r1 - FIX(0.41869) * g1 -
297 FIX(0.08131) * b1 + 4 * ONE_HALF - 1) >> (SCALEBITS + 2)) + 128;
309 /* XXX: use generic filter ? */
311 static void shrink2(UINT8 *dst, int dst_wrap,
312 UINT8 *src, int src_wrap,
313 int width, int height)
318 for(;height > 0; height--) {
322 for(w = width;w >= 4; w-=4) {
323 d[0] = (s1[0] + s2[0]) >> 1;
324 d[1] = (s1[1] + s2[1]) >> 1;
325 d[2] = (s1[2] + s2[2]) >> 1;
326 d[3] = (s1[3] + s2[3]) >> 1;
332 d[0] = (s1[0] + s2[0]) >> 1;
343 static void shrink22(UINT8 *dst, int dst_wrap,
344 UINT8 *src, int src_wrap,
345 int width, int height)
350 for(;height > 0; height--) {
354 for(w = width;w >= 4; w-=4) {
355 d[0] = (s1[0] + s1[1] + s2[0] + s2[1] + 2) >> 1;
356 d[1] = (s1[2] + s1[3] + s2[2] + s2[3] + 2) >> 1;
357 d[2] = (s1[4] + s1[5] + s2[4] + s2[5] + 2) >> 1;
358 d[3] = (s1[6] + s1[7] + s2[6] + s2[7] + 2) >> 1;
364 d[0] = (s1[0] + s1[1] + s2[0] + s2[1] + 2) >> 1;
375 static void grow22(UINT8 *dst, int dst_wrap,
376 UINT8 *src, int src_wrap,
377 int width, int height)
382 for(;height > 0; height--) {
385 for(w = width;w >= 4; w-=4) {
402 static void img_copy(UINT8 *dst, int dst_wrap,
403 UINT8 *src, int src_wrap,
404 int width, int height)
406 for(;height > 0; height--) {
407 memcpy(dst, src, width);
413 #define SCALE_BITS 10
415 #define C_Y (76309 >> (16 - SCALE_BITS))
416 #define C_RV (117504 >> (16 - SCALE_BITS))
417 #define C_BU (138453 >> (16 - SCALE_BITS))
418 #define C_GU (13954 >> (16 - SCALE_BITS))
419 #define C_GV (34903 >> (16 - SCALE_BITS))
421 #define RGBOUT(r, g, b, y1)\
423 y = (y1 - 16) * C_Y;\
424 r = cm[(y + r_add) >> SCALE_BITS];\
425 g = cm[(y + g_add) >> SCALE_BITS];\
426 b = cm[(y + b_add) >> SCALE_BITS];\
429 /* XXX: no chroma interpolating is done */
430 static void yuv420p_to_bgra32(AVPicture *dst, AVPicture *src,
431 int width, int height)
433 UINT8 *y1_ptr, *y2_ptr, *cb_ptr, *cr_ptr, *d, *d1, *d2;
434 int w, y, cb, cr, r_add, g_add, b_add, width2;
435 UINT8 *cm = cropTbl + MAX_NEG_CROP;
438 y1_ptr = src->data[0];
439 cb_ptr = src->data[1];
440 cr_ptr = src->data[2];
442 for(;height > 0; height -= 2) {
444 d2 = d + dst->linesize[0];
445 y2_ptr = y1_ptr + src->linesize[0];
446 for(w = width2; w > 0; w --) {
447 cb = cb_ptr[0] - 128;
448 cr = cr_ptr[0] - 128;
449 r_add = C_RV * cr + (1 << (SCALE_BITS - 1));
450 g_add = - C_GU * cb - C_GV * cr + (1 << (SCALE_BITS - 1));
451 b_add = C_BU * cb + (1 << (SCALE_BITS - 1));
453 /* output 4 pixels */
454 RGBOUT(d1[2], d1[1], d1[0], y1_ptr[0]);
455 RGBOUT(d1[6], d1[5], d1[4], y1_ptr[1]);
456 RGBOUT(d2[2], d2[1], d2[0], y2_ptr[0]);
457 RGBOUT(d2[6], d2[5], d2[4], y2_ptr[1]);
459 d1[3] = d1[7] = d2[3] = d2[7] = 255;
468 d += 2 * dst->linesize[0];
469 y1_ptr += 2 * src->linesize[0] - width;
470 cb_ptr += src->linesize[1] - width2;
471 cr_ptr += src->linesize[2] - width2;
475 /* XXX: no chroma interpolating is done */
476 static void yuv420p_to_rgba32(AVPicture *dst, AVPicture *src,
477 int width, int height)
479 UINT8 *y1_ptr, *y2_ptr, *cb_ptr, *cr_ptr, *d, *d1, *d2;
480 int w, y, cb, cr, r_add, g_add, b_add, width2;
481 UINT8 *cm = cropTbl + MAX_NEG_CROP;
484 y1_ptr = src->data[0];
485 cb_ptr = src->data[1];
486 cr_ptr = src->data[2];
488 for(;height > 0; height -= 2) {
490 d2 = d + dst->linesize[0];
491 y2_ptr = y1_ptr + src->linesize[0];
492 for(w = width2; w > 0; w --) {
493 cb = cb_ptr[0] - 128;
494 cr = cr_ptr[0] - 128;
495 r_add = C_RV * cr + (1 << (SCALE_BITS - 1));
496 g_add = - C_GU * cb - C_GV * cr + (1 << (SCALE_BITS - 1));
497 b_add = C_BU * cb + (1 << (SCALE_BITS - 1));
499 /* output 4 pixels */
500 RGBOUT(d1[0], d1[1], d1[2], y1_ptr[0]);
501 RGBOUT(d1[4], d1[5], d1[6], y1_ptr[1]);
502 RGBOUT(d2[0], d2[1], d2[2], y2_ptr[0]);
503 RGBOUT(d2[4], d2[5], d2[6], y2_ptr[1]);
505 d1[3] = d1[7] = d2[3] = d2[7] = 255;
514 d += 2 * dst->linesize[0];
515 y1_ptr += 2 * src->linesize[0] - width;
516 cb_ptr += src->linesize[1] - width2;
517 cr_ptr += src->linesize[2] - width2;
521 /* XXX: no chroma interpolating is done */
522 static void yuv420p_to_rgb24(AVPicture *dst, AVPicture *src,
523 int width, int height)
525 UINT8 *y1_ptr, *y2_ptr, *cb_ptr, *cr_ptr, *d, *d1, *d2;
526 int w, y, cb, cr, r_add, g_add, b_add, width2;
527 UINT8 *cm = cropTbl + MAX_NEG_CROP;
530 y1_ptr = src->data[0];
531 cb_ptr = src->data[1];
532 cr_ptr = src->data[2];
534 for(;height > 0; height -= 2) {
536 d2 = d + dst->linesize[0];
537 y2_ptr = y1_ptr + src->linesize[0];
538 for(w = width2; w > 0; w --) {
539 cb = cb_ptr[0] - 128;
540 cr = cr_ptr[0] - 128;
541 r_add = C_RV * cr + (1 << (SCALE_BITS - 1));
542 g_add = - C_GU * cb - C_GV * cr + (1 << (SCALE_BITS - 1));
543 b_add = C_BU * cb + (1 << (SCALE_BITS - 1));
545 /* output 4 pixels */
546 RGBOUT(d1[0], d1[1], d1[2], y1_ptr[0]);
547 RGBOUT(d1[3], d1[4], d1[5], y1_ptr[1]);
548 RGBOUT(d2[0], d2[1], d2[2], y2_ptr[0]);
549 RGBOUT(d2[3], d2[4], d2[5], y2_ptr[1]);
558 d += 2 * dst->linesize[0];
559 y1_ptr += 2 * src->linesize[0] - width;
560 cb_ptr += src->linesize[1] - width2;
561 cr_ptr += src->linesize[2] - width2;
565 /* XXX: no chroma interpolating is done */
566 static void yuv422p_to_rgb24(AVPicture *dst, AVPicture *src,
567 int width, int height)
569 UINT8 *y1_ptr, *cb_ptr, *cr_ptr, *d, *d1;
570 int w, y, cb, cr, r_add, g_add, b_add, width2;
571 UINT8 *cm = cropTbl + MAX_NEG_CROP;
574 y1_ptr = src->data[0];
575 cb_ptr = src->data[1];
576 cr_ptr = src->data[2];
578 for(;height > 0; height --) {
580 for(w = width2; w > 0; w --) {
581 cb = cb_ptr[0] - 128;
582 cr = cr_ptr[0] - 128;
583 r_add = C_RV * cr + (1 << (SCALE_BITS - 1));
584 g_add = - C_GU * cb - C_GV * cr + (1 << (SCALE_BITS - 1));
585 b_add = C_BU * cb + (1 << (SCALE_BITS - 1));
587 /* output 2 pixels */
588 RGBOUT(d1[0], d1[1], d1[2], y1_ptr[0]);
589 RGBOUT(d1[3], d1[4], d1[5], y1_ptr[1]);
596 d += dst->linesize[0];
597 y1_ptr += src->linesize[0] - width;
598 cb_ptr += src->linesize[1] - width2;
599 cr_ptr += src->linesize[2] - width2;
603 /* XXX: always use linesize. Return -1 if not supported */
604 int img_convert(AVPicture *dst, int dst_pix_fmt,
605 AVPicture *src, int pix_fmt,
606 int width, int height)
610 assert(pix_fmt != PIX_FMT_ANY && dst_pix_fmt != PIX_FMT_ANY);
612 if (dst_pix_fmt == pix_fmt) {
614 case PIX_FMT_YUV420P:
620 img_copy(dst->data[i], dst->linesize[i],
621 src->data[i], src->linesize[i],
628 } else if (dst_pix_fmt == PIX_FMT_YUV420P) {
631 case PIX_FMT_YUV410P:
632 img_copy(dst->data[0], dst->linesize[0],
633 src->data[0], src->linesize[0],
635 grow22(dst->data[1], dst->linesize[1],
636 src->data[1], src->linesize[1],
638 grow22(dst->data[2], dst->linesize[2],
639 src->data[2], src->linesize[2],
642 case PIX_FMT_YUV420P:
644 img_copy(dst->data[i], dst->linesize[i],
645 src->data[i], src->linesize[i],
649 case PIX_FMT_YUV422P:
650 img_copy(dst->data[0], dst->linesize[0],
651 src->data[0], src->linesize[0],
656 shrink2(dst->data[i], dst->linesize[i],
657 src->data[i], src->linesize[i],
661 case PIX_FMT_YUV444P:
662 img_copy(dst->data[0], dst->linesize[0],
663 src->data[0], src->linesize[0],
668 shrink22(dst->data[i], dst->linesize[i],
669 src->data[i], src->linesize[i],
674 yuv422_to_yuv420p(dst->data[0], dst->data[1], dst->data[2],
675 src->data[0], width, height);
678 rgb24_to_yuv420p(dst->data[0], dst->data[1], dst->data[2],
679 src->data[0], width, height);
682 rgba32_to_yuv420p(dst->data[0], dst->data[1], dst->data[2],
683 src->data[0], width, height);
686 bgr24_to_yuv420p(dst->data[0], dst->data[1], dst->data[2],
687 src->data[0], width, height);
690 bgra32_to_yuv420p(dst->data[0], dst->data[1], dst->data[2],
691 src->data[0], width, height);
696 } else if (dst_pix_fmt == PIX_FMT_RGB24) {
698 case PIX_FMT_YUV420P:
699 yuv420p_to_rgb24(dst, src, width, height);
701 case PIX_FMT_YUV422P:
702 yuv422p_to_rgb24(dst, src, width, height);
707 } else if (dst_pix_fmt == PIX_FMT_RGBA32) {
709 case PIX_FMT_YUV420P:
710 yuv420p_to_rgba32(dst, src, width, height);
715 } else if (dst_pix_fmt == PIX_FMT_BGRA32) {
717 case PIX_FMT_YUV420P:
718 yuv420p_to_bgra32(dst, src, width, height);
729 /* filter parameters: [-1 4 2 4 -1] // 8 */
730 static void deinterlace_line(UINT8 *dst, UINT8 *src, int src_wrap,
733 UINT8 *cm = cropTbl + MAX_NEG_CROP;
737 for(;size > 0;size--) {
748 dst[0] = cm[(sum + 4) >> 3];
754 /* deinterlacing : 2 temporal taps, 3 spatial taps linear filter. The
755 top field is copied as is, but the bottom field is deinterlaced
756 against the top field. */
757 static void deinterlace_bottom_field(UINT8 *dst, int dst_wrap,
758 UINT8 *src1, int src_wrap,
759 int width, int height)
765 buf = (UINT8*)av_malloc(5 * width);
768 for(y=0;y<height;y+=2) {
769 /* copy top field line */
770 memcpy(dst, src, width);
772 src += (1 - 2) * src_wrap;
774 if (y1 >= 0 && (y1 + 4) < height) {
775 /* fast case : no edges */
776 deinterlace_line(dst, src, src_wrap, width);
778 /* in order to use the same function, we use an intermediate buffer */
782 memcpy(ptr, src1, width);
783 else if (y1 >= height)
784 memcpy(ptr, src1 + (height - 1) * src_wrap, width);
786 memcpy(ptr, src1 + y1 * src_wrap, width);
790 deinterlace_line(dst, buf, width, width);
793 src += (2 + 1) * src_wrap;
799 /* deinterlace, return -1 if format not handled */
800 int avpicture_deinterlace(AVPicture *dst, AVPicture *src,
801 int pix_fmt, int width, int height)
805 if (pix_fmt != PIX_FMT_YUV420P &&
806 pix_fmt != PIX_FMT_YUV422P &&
807 pix_fmt != PIX_FMT_YUV444P)
809 if ((width & 1) != 0 || (height & 3) != 0)
815 case PIX_FMT_YUV420P:
819 case PIX_FMT_YUV422P:
826 deinterlace_bottom_field(dst->data[i], dst->linesize[i],
827 src->data[i], src->linesize[i],